Part 4 — Medical management
Chapter 12
Smart Glucose Monitoring
Imagine trying to understand a film from a single photograph. That is what one fingerstick reading is like: a sharp, accurate picture of one moment, with no way to tell whether glucose was climbing, falling, or holding steady when the shutter clicked. A glucose sensor worn on the skin turns the photograph into a film. It shows what happened after breakfast, during a walk, and while you slept, and it shows patterns that no clinic visit or HbA1c can reveal.

Technology has moved quickly, and it is not equally available everywhere. In some countries a sensor is prescribed at diagnosis; in others a box of test strips is a monthly expense weighed against food and rent. This chapter is written for both. It also aims to keep you from expecting too much: a smartwatch cannot yet read your glucose through the skin, no algorithm can tell you what to do about your medicines, and more data is not automatically better care.

Section 12.1 explains how meters and sensors work, who benefits, what they cost, and what to do if a sensor is out of reach. Section 12.2 teaches you to read the data and to test your own response to a meal. Section 12.3 looks beyond the sensor at pumps, connected pens, apps, artificial intelligence and "non-invasive" devices, and ends with questions to ask before you choose.

12.1 Meters and Sensors: What Each One Tells You

The fingerstick meter

A blood glucose meter (BGM) measures glucose in a drop of capillary blood, the blood in the tiny vessels of the fingertip, using an enzyme-coated test strip. It shows one moment, and only when you choose to look.

The international standard for home meters (ISO 15197:2013) requires that more than 95 percent of results fall within 15 mg/dL of a laboratory reference below 100 mg/dL, and within 15 percent at 100 mg/dL or above.[@jeon2016] So a reading of 140 mg/dL (7.8 mmol/L) can correspond to a true value about 20 mg/dL higher or lower and still meet the standard (15 percent of 140, calculated): good enough for daily decisions, not for arguing over ten points.

Ordinary things spoil readings. Expired or badly stored strips (a risk with second-hand strips) can give wrong results, and control solution lets you check the meter and strips. Forearm or palm readings can be less accurate than the fingertip when glucose is changing quickly, so the US Food and Drug Administration (FDA) advises the fingertip if you think you are low, do not feel your lows, or the reading does not match how you feel.[@fda2021meters] Medicines and other substances can also interfere with meter readings.[@ada2026s7] Your meter's leaflet lists other limits, such as extreme heat, cold and high altitude, and washing and drying your hands first avoids falsely high readings from food on the fingertip.

The continuous glucose monitor

A continuous glucose monitor (CGM) has a tiny filament under the skin, usually on the back of the upper arm or the abdomen, held in place by a patch. It measures glucose in the fluid between your cells (interstitial fluid) and sends a reading every few minutes to a phone, a reader, or a pump.

Because glucose has to travel from blood into that fluid, sensor readings follow blood glucose with a short delay. The ADA notes that sensor glucose correlates well with blood glucose, "although at times, it can lag if glucose levels are rising or falling rapidly".[@ada2026s7] Older sensors tested against blood samples in hospital showed delays of about 4 to 8 minutes in adolescents and 8 to 11 minutes in adults; newer sensors were not tested in that study.[@sinha2017]

{{fig:F12-A}}

Science Corner: Why can a sensor and a meter disagree? Two things are usually at work. First, the lag: during a rapid rise after a meal the fingerstick shows the higher, more current number, and the sensor catches up minutes later. Second, the sensor sits in tissue that can be pressed: in healthy volunteers, individual sensors sometimes gave readings more than 25 mg/dL from the median, mostly sudden falls, when the person lay directly on them.[@mensh2013] These "compression lows" return in section 12.2. The ADA advises a fingerstick to confirm any sensor reading that does not match how you feel.[@ada2026s7]

Types of sensor

Historically there were two families: real-time CGM (readings sent automatically, with alarms) and intermittently scanned or "flash" CGM, where you waved a reader over the sensor. The ADA now notes that most prescribed devices have converged: no swiping, a continuous stream of data, and adjustable alarms.[@ada2026s7] Professional CGM is clinic-owned, worn for a period, with data that may be blinded (you cannot see it), used to find patterns of highs and lows.[@ada2026s7] Implantable sensors, placed under the skin by a clinician, last for months and are an option in some countries. None replaces the meter: people who use a CGM must always have access to fingerstick testing.[@ada2026s7]

Who benefits, according to the guidelines

The ADA recommends CGM from diagnosis, and at any time afterwards, for children, adolescents and adults who use insulin (the strongest evidence grade), for people who take other medicines that can cause low glucose, and for anyone in whom a sensor helps with management.[@ada2026s7] For people with type 2 diabetes who do not use insulin, the ADA notes that randomised trials are increasing and have generally shown greater benefits from CGM than from meters for HbA1c and time in range, along with greater satisfaction.[@ada2026s7] A 2026 meta-analysis of nine trials in 1,453 adults with type 2 diabetes found that CGM may improve treatment satisfaction, but that its effects on diabetes distress, well-being and quality of life remain unclear.[@zhang2026cgm]

National guidance is narrower. A 2022 review found that the ADA, NICE and Diabetes Canada all recommended real-time CGM for adults with type 1 diabetes.[@cadth2022] For adults with type 2 diabetes on multiple daily insulin injections, NICE's quality standard says CGM should be offered to those with a condition or disability that means they cannot use fingerstick testing.[@nice2023qs209] In 2024 the FDA cleared the first over-the-counter CGM for adults who do not use insulin; it is not designed for people with problematic hypoglycaemia, and users are asked to speak to a healthcare provider before making medical decisions from it.[@fda2024otc]

Cost and access around the world

Access depends far more on where you live than on what the guidelines say. The World Health Organization added personal-use glucose meters to its Essential Diagnostics List for the first time in 2023.[@who2023eid] CGM is a different story. A 2026 review in Africa reported system costs of US$2,000 to 6,000 a year, above per-capita income in many countries; in a multi-country survey, people considered US$19 to 26 a month affordable against a market cost of about US$120, and in Nigeria more than 95 percent of treatment costs are paid by patients.[@olamoyegun2026] Even in wealthy countries, the ADA acknowledges that insurance coverage can lag behind the technology.[@ada2026s7]

If a personal sensor is out of reach, three approaches are supported by the guidelines.

12.2 Reading Your Data

Time in range, extended

Chapter 4 introduced time in range: the share of the day spent between 70 and 180 mg/dL (3.9 to 10.0 mmol/L), with a goal of more than 70 percent for most adults, less than 4 percent below range, and looser or stricter targets for older and higher-risk people.[@battelino2019] Three points extend that picture. First, the numbers are easier to feel as minutes: 4 percent of 24 hours is about 58 minutes and 1 percent is about 14 minutes (calculated), so a quarter of an hour below 54 mg/dL (3.0 mmol/L) already exceeds the daily allowance for very low readings. Second, pregnancy uses a narrower band: for pregnancy with type 1 diabetes the consensus range is 63 to 140 mg/dL (3.5 to 7.8 mmol/L), and your team sets the targets that apply to you.[@ada2019pr] Third, the target is not a description of normal. In 153 healthy people without diabetes wearing sensors for up to 10 days, average glucose was about 98 to 99 mg/dL (5.4 to 5.5 mmol/L), and the median time between 70 and 140 mg/dL was 96 percent.[@shah2019] A time in range of 70 percent on the 70 to 180 scale is a realistic goal for people with diabetes, not a failure to be "normal".

Why does time in range matter? In a re-analysis of the DCCT trial in type 1 diabetes, each 10 percentage points lower time in range was associated with a 64 percent higher hazard of retinopathy progression and a 40 percent higher hazard of microalbuminuria, and time in range correlated with HbA1c at -0.79.[@beck2019tir] This is an association in type 1 diabetes, but it explains why time in range is treated as a companion to HbA1c.

{{fig:F12-B}}

When the sensor's HbA1c estimate and the laboratory disagree

Chapter 4 mentioned the glucose management indicator (GMI), an HbA1c-like value calculated from sensor data. It will not always match a laboratory HbA1c. Across studies, 26 percent to 68 percent of people differ by 0.5 percentage points or more, for reasons that include differences between sensors, the fact that HbA1c reflects two to three months while a sensor covers a shorter period, and non-glucose factors such as red blood cell lifespan and genetics.[@selvin2024] Neither number is "the truth". If they disagree, look at the whole sensor report and talk to your team. Chapter 4 explains when HbA1c itself can mislead.

Trend arrows and alarms

Most sensors show an arrow beside the number to say how fast glucose is changing, and can sound alarms for lows, highs, or a predicted low. The number is where you are; the arrow is the direction of travel, and it matters most when glucose is changing quickly, which is when the sensor lags.

This book does not tell you what to do with an arrow or an alarm; that depends on your medicines, targets and your team's plan. Follow your care team's plan, and if you do not have one, ask for one. Alarms also have a cost: a review in children and adolescents with type 1 diabetes defined alarm fatigue as reduced responsiveness caused by frequent false alarms, listed compression artefacts among the causes, and reported sleep disruption, recommending individualised alarm settings, education and psychological support.[@giza2025] The evidence is from young people, but an alarm that constantly interrupts you is worth reviewing with your team.

Night, dawn, and false lows

Chapter 10 described the dawn phenomenon and the debated rebound after a night-time low; a sensor is where you can see them. In 248 adults with type 2 diabetes who were not on insulin, the median dawn rise was about 16 mg/dL, and its estimated effect on HbA1c averaged 0.39 percentage points.[@monnier2013] In 2,600 people with type 2 diabetes on stable insulin, morning glucose was lower after a night-time low, not higher, and the authors found no support for a rebound effect.[@huang2022] In type 1 diabetes, however, 248 of 755 sensor users (32.8 percent) had a night-time low followed by a high before 6 a.m., so the pattern can be real for some people.[@gonzalezvidal2025] CGM can also reveal lows that cause no symptoms.[@ada2026s6]

Night data needs one caution. A sudden low that begins when you roll onto the sensor and recovers when you change position may be a compression low, with the sensor under-reading.[@mensh2013] If an alarm wakes you with a low reading and you feel unwell or are unsure, follow your care team's plan and check with a fingerstick (Chapter 13 covers treating lows). Do not change your evening medicine or food because of one strange night; look for a pattern over several nights and discuss it with your team.

Test your own response to a meal, a meal order, or a walk

Chapters 5, 8 and 9 said that the table or the study is where you start and that your own meter or sensor gives the final answer. Even the same person can respond differently to the same food on different days (Chapter 5).[@vegalopez2007] The "test one meal twice" method is a simple way to learn something useful despite that noise.

  1. Choose one ordinary meal you eat often, at a usual time of day, and one thing to compare: the order in which you eat (vegetables and protein before the starch), the portion of the starch, or a walk of your usual length afterwards.
  2. Day one: eat the meal your usual way. Test before eating, and again at about one hour and two hours after the start of the meal. (For most non-pregnant adults, the ADA measures the peak after-meal glucose 1 to 2 hours after the start of the meal, with a target below 180 mg/dL.[@ada2023targets]) With a sensor, mark the meal time and read the curve.
  3. Day two: repeat the same meal at the same time, changing only the one thing. Try to match sleep, activity, and stress as far as you can.
  4. Compare the two curves or sets of numbers, and repeat the pair on another day. A difference that shows up twice deserves more trust than one that shows up once.

{{fig:F12-C}}

This is a way of gathering information, not of passing a test; you are learning what works for your body. One pair of tests can mislead, because stress, poor sleep, a cold, or a change in medicine can all shift a reading (Chapter 10). If you use insulin or medicines that can cause lows, changing meal order, portion, or timing can alter the risk of hypoglycaemia, so talk to your care team first, and never use the results to change a dose on your own.

When readings mislead: substances, skin, and privacy

Interfering substances. Some substances make certain sensors read higher than the true glucose. The ADA lists high-dose acetaminophen (paracetamol; more than 4 g a day for two of the systems listed, and any dose for another), vitamin C (more than 500 mg a day for two systems and more than 1,000 mg a day for two others), and hydroxyurea, a medicine used mainly in blood disorders.[@ada2026s7] Thresholds differ by device and change with new models, so tell your team about your medicines and supplements, read your sensor's leaflet, and use a fingerstick when a reading seems out of line.

Skin reactions. Adhesive patches can cause irritation or allergy, and the ADA asks that skin reactions be assessed so that people can keep using their device.[@ada2026s7] A 2025 review reported confirmed allergic contact dermatitis in 3.8 percent of 1,036 sensor users in one Belgian study, and 0.8 to 1.0 percent in Finnish data; reactions often appeared five to seven months after starting, and protective creams, sprays, and dressings had limited success.[@degroot2025] If you develop a rash, tell your team.

Data sharing and privacy. Sensor data can be shared with family, a clinic, or an app. The ADA advises reading a company's data privacy and sharing policy before you enter data into an app, and notes that established cloud services designed for clinical data have security features and comply with US health-privacy law.[@ada2026s7] Decide who sees your numbers, and for what.

12.3 Technology Beyond the Sensor

Insulin pumps and automated insulin delivery

An insulin pump is a small device that delivers insulin continuously through a tube or a patch placed on the skin. In an automated insulin delivery (AID) system, sometimes called a hybrid closed loop or "artificial pancreas", the pump is linked to a CGM and an algorithm adjusts background insulin from moment to moment. It is "hybrid" because the user still tells the system about meals. The algorithm is part of a regulated device set up by your diabetes team.

For type 1 diabetes, the evidence is substantial. A meta-analysis of 22 randomised trials in 2,376 people found time in range 10.87 percentage points higher with AID than with usual care, HbA1c 0.37 points lower, less nocturnal hypoglycaemia, and no increase in diabetic ketoacidosis or severe hypoglycaemia.[@godoi2023] A 2025 meta-analysis of 65 trials in 3,623 people found a time-in-range gain of 11.74 percentage points (95% confidence interval 9.37 to 14.12) and about 1.2 percentage points less time below range.[@fan2025] In the UK, NICE recommends hybrid closed loop as an option for adults with type 1 diabetes whose HbA1c is 58 mmol/mol (7.5 percent) or more, or who have disabling hypoglycaemia, despite the best possible management, and for children and for pregnancy.[@nice2023hcl]

For type 2 diabetes the evidence is newer. In a 13-week trial of 319 adults with insulin-treated type 2 diabetes, HbA1c fell by 0.9 percentage points with AID and by 0.3 with standard insulin delivery plus a CGM, and time in range rose from 48 to 64 percent.[@iqp2025] A meta-analysis pooling 10 randomised trials (673 people) with single-arm and real-world studies found about 16 percentage points more time in range (roughly 229 minutes a day), no significant difference in time below range, and rare serious adverse events, though device problems were slightly more common and the improvements were larger in trials than in everyday practice.[@zeng2025] The ADA now prefers AID to injections or a standard pump in type 1 diabetes and adults with type 2 diabetes, tailored to individual needs and preferences.[@ada2026s7]

Science Corner: What a "closed loop" does and does not do A hybrid closed loop watches your sensor and adjusts background insulin toward a target. It cannot see what you are about to eat, when you will exercise, or that your sensor is under-reading because you slept on it, so you remain responsible for telling it about meals and for checking a reading that does not make sense. Pumps depend on supplies and training, and if delivery fails, glucose can rise quickly, so trials track device problems and ketoacidosis as safety outcomes. These are matters for your diabetes team and for Chapter 13.

Smart pens, apps, and telehealth

A connected ("smart") insulin pen or pen cap records the time of each injection and sends it to an app, where it can be read alongside sensor data. The ADA recommends offering connected pens to people on multiple daily injections when appropriate, asks that standardised one-page reports be available for all sensor, pump and connected devices, and suggests combining technology with online or virtual licensed coaching.[@ada2026s7] Telehealth has particular promise where clinics are far away, since sensor data can be reviewed remotely. Some apps include insulin dose calculators; these are medical software for people whose team has trained them to use them, and this book gives no dosing rules.

Artificial intelligence and predictive tools

Some tools already use algorithms, such as a sensor that warns of a low before it happens. In Chapter 5 we met a study in which sensor data from 800 adults, combined with other information, were used to build a model that predicted glucose responses to meals better than carbohydrate counting alone (a correlation of 0.68 versus 0.38).[@zeevi2015] The interest in personalised, data-driven food advice comes from this kind of work. Two cautions apply. The model in that study was a research tool, and this chapter cites no trial showing that consumer "AI" apps improve long-term outcomes such as HbA1c or complications. And many apps for people without diabetes are sold as wellness products, not medical devices, so treat their advice like any general nutrition advice.

Smartwatches, rings, and "non-invasive" glucose

There is no shortcut yet. On 21 February 2024 the FDA issued a safety communication stating that it "has not authorized, cleared, or approved any smartwatch or smart ring that is intended to measure or estimate blood glucose values on its own". It warned that inaccurate readings could lead to wrong medicine decisions and dangerous lows, and advised people not to buy or use smartwatches or rings that claim to measure glucose. Only authorised CGMs and meters that pierce the skin are cleared for that purpose.[@fda2024smart] The research picture matches. A 2025 review of optical, electrical, and antenna-based methods described promising laboratory results, but noted that performance of some approaches became much more variable in real-life conditions.[@shi2025] A watch can count steps and sleep, which Chapters 9 and 10 value, but a glucose number from a watch should not be used for any medical decision.

Avoiding data overwhelm and sensor anxiety

Seeing every rise after a meal can produce guilt or worry, and constant alarms can wear you down. The evidence on emotional effects is mixed: the type 2 meta-analysis above could not say whether CGM reduces diabetes distress,[@zhang2026cgm] and alarm fatigue is a recognised problem.[@giza2025] A few habits help.

If numbers begin to dominate your day, that is common and understandable, and it deserves support; Chapter 16 covers diabetes distress and burnout.

Questions to ask when choosing a device

Key Takeaways

Action Points

  1. Check your meter. Look at the expiry date on your strips and how they have been stored, and ask your team or pharmacist how to check the meter with control solution.
  2. Ask about your monitoring plan. Ask your care team what you should measure, when, and what to do with the result, and whether a sensor, a borrowed sensor, or structured testing suits you.
  3. Look at one report. If you have a sensor, request your time in range, time below range, and overnight pattern, and pick two numbers to follow.
  4. Test one meal twice. With your team's agreement, choose one meal and one change, follow the four steps in section 12.2, and write down what you see on two separate days.
  5. Choose your alarms and your privacy settings. Agree which alarms you need, and decide who can see your data.

This book is intended for education and does not replace personal medical advice. If you have diabetes or take glucose-lowering medication, please consult your healthcare team before changing your diet, exercise, or treatment.

All chapters